What We Do

We prove what AI can actually absorb.

Not a pilot. Not a demo. A 90-day embedded engagement that produces a financial model your board can fund — built on your data, under a protocol designed so we cannot grade our own homework.

The core service

The Maintenance Tax Assessment

We embed a Forward Deployed Engineer inside your engineering organisation. They work in your systems, with your team, against your actual 24 months of maintenance history — not a sanitised extract in a separate environment.

At the end of 90 days, you have three things:

An evidence base

Every issue class scored under the blind protocol. AI model answer frozen before it sees the fix your engineers shipped. Every record auditable by your own team.

A financial model

Built on your numbers, not our assumptions. Cost recaptured, capacity redirected, payback and NPV — in three scenarios, labelled by source.

An operating blueprint

A function specification for the team that would run this — governance model, rollback posture, SLA structure — designed to be presented without us in the room.

An honest conclusion

If the evidence does not justify the investment, the report says so. That is not a failure condition — it is the point. An assessment that can only conclude yes is not an assessment.

Engagement structure

Four properties that make the output fundable.

Blind by design

Our AI model never sees the fix your engineers shipped until after its own answer is frozen and timestamped. The first question a CTO asks is whether we saw the answer first. We built the method so the answer is no.

Every number sourced

Each figure is labelled: your data, or our measurement. Never our assumption. A CFO who can trace every input trusts the model. That traceability is built into the format, not added at the end.

One dataset, three lenses

The same evidence answers your CFO, your CPO, and your CTO. The numbers cannot contradict each other because there is only one dataset underneath all three outputs.

Audit it yourself

We invite your engineers to independently score a random sample of our findings. We offer this before anyone asks — because the value of the method is that it survives scrutiny.

The return

The saving is not the prize. The capacity is.

A cost story ends at the saving. This one starts there. Every hour recovered from maintenance is an hour of already-funded engineering capacity that can be pointed at the roadmap — new product, new features, the work that shows up in revenue rather than in run-rate.

This is why the framing is redirection, not reduction. A maintenance-automation programme sold internally as a headcount exercise loses the engineers it needs to succeed. Sold as freeing your best people to build, it is one of the few AI initiatives an engineering organisation will actively pull for.

A 30-minute call tells us whether there is a question worth answering. If there is not, we will say so.

Start a conversation →